./whisper-base-ea_base

This model is a fine-tuned version of openai/whisper-base on the Afrispeech-200 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6880
• Wer Ortho: 0.2755
• Wer: 0.2202
• Cer: 0.0998
• Precision: 0.8628
• Recall: 0.8622
• F1: 0.8616

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer	Cer	Precision	Recall	F1
0.9241	0.4237	100	0.8794	0.2973	0.2468	0.1060	0.8428	0.8469	0.8443
0.7528	0.8475	200	0.7464	0.2903	0.2354	0.1032	0.8583	0.8593	0.8581
0.5275	1.2712	300	0.7158	0.2778	0.2285	0.1000	0.8619	0.8627	0.8616
0.5686	1.6949	400	0.6956	0.2805	0.2255	0.1021	0.8638	0.8632	0.8626
0.3472	2.1186	500	0.6880	0.2755	0.2202	0.0998	0.8628	0.8622	0.8616

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1